In the past, analog to digital converters using recirculation of the remainder systems were difficult and expensive to build as production devices. The converters required many components and because of this the tolerance variations in the components caused accurate systems to be extremely expensive. Various components would have to be matched to reduce the affect of tolerance buildup and components which could not be matched would have to be manually adjusted by skilled technicians. Other components were required which were very accurate and had minimal tolerance variations from component to component. All of these factors contributed to very expensive analog to digital converter.
In order to obtain five and a half digits accuracy, the converters required additional components above those described in the "Analog Digital Converter and Indicator Using Recirculation of Remainder" patent, U.S. Pat. No. 3,703,002. When increasing the accuracy of the Van Saun system, there are many undesirable effects such as the switch charge injection phenomenon, which can cause significant errors in high resolution and accuracy systems. With other components, for example the field effect transistor switches used in the Van Saun system, it was discovered that temperature variations during the use of the converter would introduce sufficient error to render the final systems inaccurate.
In the case of amplifiers, to obtain high accuracy, the amplifiers had to be individually manually adjusted, to customize the band width or the frequency response. To achieve a desired transient response, this manual adjustment meant, in some cases, adding or deleting capacitors while observing a waveform on an oscilloscope to achieve a specified response.
It further developed, that the previous methods of avoiding common mode rejection errors were insufficient to obtain high accuracies at low cost. When integrating the amplifiers, the theoretical common mode rejection ratio of 120 db was found not to be obtainable with conventional CMOS integrated circuit technology. Although computer analysis and simulations indicated the feasibility of amplifiers with 110 db common mode rejection ratio, in production, even in closely monitored production manufacturing situations, it was not possible to repeatedly obtain even acceptable yields of devices with such common mode rejection ratios.
With extensive experimentation, it was determined that the direction of current flow through a bilateral switch had a significant effect at the high levels of accuracy desired. The switches typically also had a temperature coefficient of one-half to six-tenths of a percent per degree centigrade, and this was sufficient to introduce inaccuracies in a device which had to work over a broad temperature range from zero to seventy degrees centigrade.
In examining the prior recirculation of remainder systems, it was determined that several follower amplifiers were required to charge the storage capacitors. To reduce cost, it is necessary to eliminate some of the follower amplifiers. Attempts at having a regular operational amplifier function as a follower amplifier were tried unsuccessfully.
Another problem with the prior art was that the existing amplifiers were not fast enough and to increase speed it was necessary to add additional amplifiers which resulted in the need for additional compensation to stabilize the amplifiers.
The additional amplifiers created additional problems in zeroing the converter circuitry. The various systems used in the past tended to have a multitude of offset errors due to the multitude of amplifiers. This made calibration to determine the zero values of the circuitry very difficult and time consuming.
The above problems appeared to be unsolvable and the technology was not advancing until a fortuitous accident. While one of a number of experimental breadboards was being tested, a wire broke. The breadboard with the broken wire had a huge offset, but the linearity was excellent. This resulted in the hindsight realization that unidirectional switches would solve the switch problem and that changing the reference voltages on the A/D circuit would eliminate the amplifier problems. This eventually led to a simplification of the amplifier circuitry to a practical level by elimination of many components.
Evolving from the simplification effort, it was determined that one analog to digital amplifier could be made to perform several functions during different portions of the analog to digital conversion cycle. With an approach, the proper autozero technique was discovered to eliminate the huge offsets and to provide the desired output accuracy.
Still further, it was determined that prior art systems were subject to cardinal point errors where the digital outputwould undergo a discontinuity on tracking the analog input. Heretofore, this was considered an inherent aspect of the recirculating remainder analog to digital converter systems.